3D Object Classification on ModelNet40 Few-shot (5-way & 10-way)
65.41Accuracy (5-way, 10-shot)PointCNN
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PointCNNEvaluation Protocol=FSL setting, Supervision=Supervised (Random Weight Initialization)2020.09 | 65.41 | 68.64 | 46.6 | 49.95 | |
| Our+PointNetBackbone=PointNet, Pre-training=Our self-supervised point embeddings2020.09 | 63.2 | 68.9 | 49.15 | 50.1 | |
| Our+DGCNNBackbone=DGCNN, Pre-training=Our self-supervised point embeddings2020.09 | 60 | 65.7 | 48.5 | 53 | |
| 3D-GANEvaluation Protocol=Linear SVM classification on final embeddings, Supervision=Unsupervised2020.09 | 55.8 | 65.8 | 40.25 | 48.35 | |
| PointNetEvaluation Protocol=FSL setting, Supervision=Supervised (Random Weight Initialization)2020.09 | 51.97 | 57.81 | 46.6 | 35.2 | |
| PointCapsNetEvaluation Protocol=Linear SVM classification on final embeddings, Supervision=Unsupervised2020.09 | 42.3 | 53 | 38 | 27.15 | |
| Latent-GANEvaluation Protocol=Linear SVM classification on final embeddings, Supervision=Unsupervised2020.09 | 41.6 | 46.2 | 32.9 | 25.45 | |
| PointNet++Evaluation Protocol=FSL setting, Supervision=Supervised (Random Weight Initialization)2020.09 | 38.53 | 42.39 | 23.05 | 18.8 | |
| Voxel+DGCNNBackbone=DGCNN, Pre-training=VoxelSSL [26]2020.09 | 34.3 | 42.2 | 26.05 | 29.9 | |
| FoldingNetEvaluation Protocol=Linear SVM classification on final embeddings, Supervision=Unsupervised2020.09 | 33.4 | 35.8 | 18.55 | 15.44 | |
| DGCNNEvaluation Protocol=FSL setting, Supervision=Supervised (Random Weight Initialization)2020.09 | 31.6 | 40.8 | 19.85 | 16.85 |